Visual Object Tracking On Davis 2017

评估指标

F-measure (Mean)

Ju0026F

Jaccard (Mean)

Params(M)

Speed (FPS)

评测结果

各个模型在此基准测试上的表现结果

						Paper Title	Repository
SAM2	-	90.7	-	224.4	-	SAM 2: Segment Anything in Images and Videos
Cutie+ (base)	93.4	90.5	87.5	17.9	-	Putting the Object Back into Video Object Segmentation
ISVOS (BL30K, MS)	93.0	89.8	86.7	-	-	Look Before You Match: Instance Understanding Matters in Video Object Segmentation	-
XMem (BL30K, MS)	92.6	89.5	86.3	-	-	XMem: Long-Term Video Object Segmentation with an Atkinson-Shiffrin Memory Model
ISVOS (MS)	-	88.6	85.8	-	-	Look Before You Match: Instance Understanding Matters in Video Object Segmentation	-
XMem (MS)	91.0	88.2	85.4	-	-	XMem: Long-Term Video Object Segmentation with an Atkinson-Shiffrin Memory Model
ISVOS (BL30K)	91.9	88.2	84.5	-	-	Look Before You Match: Instance Understanding Matters in Video Object Segmentation	-
Cutie+ (base, MEGA)	90.8	88.1	85.5	-	17.9	Putting the Object Back into Video Object Segmentation
JIMD	91.0	88.1	85.2	-	-	Memory Matching is not Enough: Jointly Improving Memory Matching and Decoding for Video Object Segmentation	-
Cutie (base)	91.1	87.9	84.6	36.4	-	Putting the Object Back into Video Object Segmentation
XMem (BL30K)	91.4	87.7	84.0	-	22.6	XMem: Long-Term Video Object Segmentation with an Atkinson-Shiffrin Memory Model
DEVA	91.0	87.6	84.2	-	25.3	Tracking Anything with Decoupled Video Segmentation
SwinB-AOTv2-L (MS)	89.8	87.0	84.2	65.6	1.3	Scalable Video Object Segmentation with Identification Mechanism
SwinB-AOST (L'=3, MS)	89.5	86.7	83.8	65.6	1.3	Scalable Video Object Segmentation with Identification Mechanism
SwinB-AOTv2-L	89.4	86.3	83.1	65.6	12.0	Scalable Video Object Segmentation with Identification Mechanism
XMem	89.5	86.2	82.9	-	22.6	XMem: Long-Term Video Object Segmentation with an Atkinson-Shiffrin Memory Model
SwinB-DeAOT-L	89.2	86.2	83.1	70.3	15.4	Decoupling Features in Hierarchical Propagation for Video Object Segmentation
RAVOS	89.3	86.1	82.9	-	42 (on 3090)	Region Aware Video Object Segmentation with Deep Motion Modeling	-
QDMN	88.6	85.6	82.5	-	-	Learning Quality-aware Dynamic Memory for Video Object Segmentation
R50-AOST (L'=3)	88.5	85.6	82.6	15.4	17.5	Scalable Video Object Segmentation with Identification Mechanism

0 of 82 row(s) selected.